A critical firmware bug in Coldcard hardware wallets allowed attackers to recover private keys offline, draining approximately $91 million in Bitcoin from roughly 4,925 addresses starting July 30, 2026. The vulnerability stemmed not from a broken cryptographic algorithm, but from a build configuration error that routed seed generation through a predictable software fallback instead of the device's hardware random number generator.

What Went Wrong Inside Coldcard's Firmware?

Coldcard, produced by Coinkite, is a cold-storage hardware wallet designed to keep cryptocurrency private keys offline and isolated from internet-connected devices. The security of any such wallet depends entirely on the unpredictability of the seed phrase generated during setup. When that randomness fails, the entire security model collapses.

The root cause was a mismatch between how the firmware was configured and how the seed-generation code actually worked. In 2021, Coldcard's developers migrated the wallet-generation process to use a function called ngu.random.bytes(). However, the build configuration disabled MicroPython's hardware random number generator (RNG) by setting a macro called MICROPY_HW_ENABLE_RNG to 0. This was intended because Coldcard had its own separate hardware RNG wrapper. But the seed-generation code didn't call that wrapper; instead, it fell back to MicroPython's software-based Yasmarang generator, which relies on predictable device metadata like the real-time clock, system timer, and device unique identifier.

The critical oversight was in how the code checked for the hardware RNG's availability. The guard statement checked whether the macro existed, not whether it was enabled. A macro defined as 0 is still technically "defined," so the build succeeded, and the seed-generation path routed through the software fallback without triggering an error.

How Did Attackers Recover Private Keys?

Because the weakness was in the seeds themselves, the attack required no on-chain exploit transaction and no vulnerable smart contract. Instead, attackers performed an offline brute-force search. For Mk2 and Mk3 hardware revisions, the compromised seed-generation process collapsed cryptographic entropy from an infeasible problem into approximately 40 bits of searchable space. For newer Mk4, Q, and Mk5 models, the entropy was higher at approximately 72 bits, but still far below the 256-bit security standard.

To put this in perspective, 40 bits of entropy means an attacker can enumerate roughly 2^40 candidate seeds, or about 1 trillion possibilities. A modern computer can test billions of candidates per second, making the entire search feasible in hours or days. Once an attacker recovered a candidate seed, they could derive the corresponding private keys and match them against known public wallet addresses on the blockchain to identify which seeds were actually in use.

As of August 7, 2026, on-chain tracking had verified a floor of approximately 1,405 Bitcoin (BTC), worth roughly $91 million at the $64,700 price that day. Wave-level attribution across ten separate theft waves reached approximately 1,433 BTC, and private-channel reconciliation with victims suggested the total could be as high as 2,055 BTC, or approximately $133 million.

Why Is This Worse Than Other Crypto Hacks?

Hardware wallets are marketed as one of the most secure ways to store cryptocurrency precisely because they keep private keys offline, away from internet-connected devices that are commonly targeted by hackers. A confirmed exploit affecting a hardware wallet challenges fundamental assumptions about cold-storage security and can erode confidence in the entire category.

The Coldcard incident contributed to a broader pattern of crypto losses in July 2026. That month saw approximately $247 million in cumulative crypto losses across all incidents, making it the second-costliest month of the year. The Coldcard exploit was a significant component of that total, though other hacks and exploits also occurred during the same period.

What Happens to Funds Generated Under the Flawed Firmware?

A critical implication for affected users is that a seed generated under the vulnerable firmware stays compromised even after Coinkite releases a patched firmware update. The vulnerability is in the seed itself, not in how the wallet uses the seed after generation. This means users who created wallets on affected Coldcard devices before the firmware fix must move their funds to a new wallet generated with the corrected firmware.

Additionally, a separate post-hotfix firmware regression was identified that can deny service before login, suggesting that the remediation process itself introduced new complications.

How to Protect Yourself If You Own a Coldcard

• Check Your Device Model and Firmware Version: Determine whether your Coldcard is a Mk2, Mk3, Mk4, Q, or Mk5 model, and verify your current firmware version. Mk2 and Mk3 devices are affected by the entropy failure, as are Mk4, Q, and Mk5 models, though with varying levels of residual entropy.
• Generate a New Seed on Patched Firmware: If you created your wallet before the firmware fix, generate a new seed phrase using the corrected firmware version. Do not rely on seeds generated under the vulnerable firmware, even if you update the device later.
• Move Funds to the New Wallet: Transfer all cryptocurrency holdings from any wallet created under the flawed firmware to a new wallet generated with the patched firmware. This is the only way to ensure your funds are not vulnerable to offline seed recovery.
• Verify Firmware Integrity: Before trusting a new seed generation, confirm that your Coldcard is running the latest patched firmware and that the hardware RNG is functioning correctly. Coinkite should provide verification tools or documentation for this purpose.
• Consider Diversifying Custody Solutions: While hardware wallets remain a best practice for self-custody, this incident highlights the importance of not relying on a single device or manufacturer. Some users may choose to split holdings across multiple hardware wallet brands or custody solutions.

What Does This Mean for Hardware Wallet Security Standards?

The Coldcard incident reveals a critical gap in how hardware wallet security is validated. The vulnerability was not in the cryptographic algorithm itself, but in a build configuration error that bypassed the intended hardware RNG. This suggests that security audits and code reviews may not catch integration failures that occur at the firmware compilation stage.

The technical details indicate that cryptographic entropy paths must fail closed, meaning they should error out if the hardware RNG is unavailable, rather than silently falling back to a weaker alternative. Additionally, entropy security must be verified end-to-end in the shipped firmware, not just at compile time.

At the broader market level, elevated monthly loss figures such as the $247 million reported for July 2026 reinforce ongoing scrutiny of crypto security practices among regulators, insurers, and institutional investors evaluating exposure to digital assets. However, the Coldcard case is particularly significant because it affects a device category that is widely recommended as a best practice for securing significant cryptocurrency holdings.

Users and institutions may respond by seeking additional verification of wallet firmware integrity, diversifying custody solutions, or awaiting detailed technical documentation from Coinkite before drawing conclusions about the safety of the device line. As additional details emerge, the crypto community will be watching for confirmation from Coinkite and independent security researchers to better understand the scope and mechanics of the exploit, as well as how it fits into the industry's broader security trends for 2026.